Rubberduck - ChatGPT for Visual Studio Code

Generates new code snippets based on natural language descriptions by sending the user's intent and current editor selection context to OpenAI's API, then inserting the generated code at the cursor position or displaying it in the sidebar. The extension reads the active editor's selected text to provide code context, enabling the model to generate syntactically appropriate code for the detected language. Generation is triggered via keyboard shortcut (Ctrl+Alt+G), command palette, or toolbar button.

Modifies selected code by sending the selection and user-provided editing instructions to OpenAI, receiving a modified version, and displaying it in a side-by-side diff viewer before applying changes. The user reviews the proposed changes and explicitly clicks 'Apply' to accept them, preventing accidental code replacement. Triggered via Ctrl+Alt+E keyboard shortcut or context menu. The diff viewer uses VS Code's native diff rendering with optional syntax highlighting toggled via the `rubberduck.syntaxHighlighting.useVisualStudioCodeColors` setting.

Provides platform-specific keyboard shortcuts for common actions (Chat, Generate Code, Edit Code) that trigger commands without opening the command palette. Shortcuts are: Chat (Ctrl+Alt+C / Ctrl+Cmd+C), Generate (Ctrl+Alt+G / Ctrl+Cmd+G), Edit (Ctrl+Alt+E / Ctrl+Cmd+E), with Windows/Linux and Mac variants. Shortcuts are customizable via VS Code's standard keybinding configuration. This enables power users to access features without mouse interaction or command palette navigation.

Analyzes selected code by sending it to OpenAI and returns a natural language explanation of what the code does, its purpose, and how it works. The explanation is displayed in the sidebar chat panel, allowing developers to understand unfamiliar code without leaving the editor. Triggered via command palette or context menu. Supports any language that VS Code can syntax-highlight, though explanation quality depends on the model's training data for that language.

Generates test code for selected code by sending it to OpenAI and returning test cases in the sidebar. The specific test framework and language are inferred from the selected code's context. Tests are displayed in the chat panel and can be copied or inserted into the editor. Implementation details of test framework selection are not documented, suggesting automatic detection based on file type or imports.

Analyzes selected code for potential bugs, security issues, or logic errors by sending it to OpenAI and returning identified problems in the sidebar chat. The analysis is performed on the selected code only, without access to the broader codebase or runtime context. Results are presented as a list of issues with explanations, allowing developers to review and decide whether to fix them.

Analyzes error messages (compiler errors, runtime exceptions, stack traces) provided by the user and returns explanations and potential fixes in the sidebar chat. The user pastes or describes the error, and OpenAI provides context about what caused it and how to resolve it. This capability bridges the gap between error output and actionable solutions without requiring manual documentation lookup.

Maintains a multi-turn conversation in the sidebar panel where users can ask questions about code, request explanations, discuss design decisions, and iterate on solutions. Each conversation thread maintains context across multiple exchanges, allowing follow-up questions and refinements. Conversations are stored in the sidebar and can be reviewed or continued later. The extension sends conversation history to OpenAI to maintain context, enabling coherent multi-turn interactions.

Allows users to create custom conversation templates that define specific prompts, instructions, or workflows for recurring tasks. Templates can be saved and reused, enabling teams to standardize how they interact with the AI for common tasks like code review, documentation, or refactoring. Implementation details are not documented, but the feature implies a template storage mechanism and template instantiation system. Users can define custom prompts that are sent to OpenAI with predefined context or instructions.

Allows users to switch between OpenAI's cloud API and a local Llama.cpp instance running CodeLlama-7B-Instruct via the `Rubberduck: Model` configuration setting. The extension abstracts the backend difference, sending the same prompts to either service and handling responses uniformly. Local mode requires users to install and configure Llama.cpp separately with the CodeLlama model (~7GB download). This capability enables offline operation and cost control at the expense of code quality and speed.

Provides a dedicated sidebar panel in VS Code where all AI interactions (code generation, explanations, tests, bug reports, conversations) are displayed and managed. The panel serves as the primary UI for Rubberduck, showing conversation threads, generated code, and explanations. The sidebar maintains chat history within the session, allowing users to scroll through previous interactions and reference earlier responses. The panel includes toolbar buttons for 'Start new chat' and 'Generate Code' for quick access to common actions.